That looks recent, so I suppose that's not the issue. Other possibilities are that the browser's VM might happen to optimize something better than LLVM's codegen. Or that emscripten libc happens to be better at something, if you use any computationally intensive method there.
In general, emscripten+asm.js are about as fast as native code. It's not too surprising to hear that sometimes a native build is 30% faster or asm.js is 20% faster. That's the range of difference between native compilers themselves anyhow. On Mon, Jul 27, 2015 at 3:43 PM, Mark Sibly <[email protected]> wrote: > I'm on a mac mini, g++ --version is: > > Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr > --with-gxx-include-dir=/usr/include/c++/4.2.1 > > Apple LLVM version 6.1.0 (clang-602.0.53) (based on LLVM 3.6.0svn) > > Target: x86_64-apple-darwin14.4.0 > > Thread model: posix > > On Tue, Jul 28, 2015 at 10:41 AM, Alon Zakai <[email protected]> wrote: > >> Which version of gcc are you using? If you're not using the latest, then >> the difference might be that emscripten does use a recent LLVM. >> >> On Mon, Jul 27, 2015 at 3:28 PM, Mark Sibly <[email protected]> >> wrote: >> >>> Ok, it appears em++ may be optimizing out some stuff that g++ isn't, as >>> the em++ version is generating less garbage than the c++ version. >>> >>> I'm using -O3 for both em++ and g++, are there any other flags I can use >>> in g++ to get similar optimizations? >>> >>> Bye, >>> Mark >>> >>> On Tuesday, July 28, 2015 at 7:53:31 AM UTC+12, Mark Sibly wrote: >>>> >>>> Hi, >>>> >>>> I am working on a 'transpiler' style language that outputs c++, which I >>>> then either compile as native code to produce desktop apps, or put through >>>> emscripten to produce web apps. >>>> >>>> I've recently been working on the garbage collection stuff, and decided >>>> to run a few speed tests as I have long been worried ask.js wouldn't be up >>>> to this sort of thing. >>>> >>>> But much to my surprise, in many cases, the asm.js code runs *faster* >>>> than the native code! I can tweak some GC settings so this isn't the case, >>>> but with default settings asm.js is about 20% faster than the corresponding >>>> native code. >>>> >>>> The test code is meaningless 'thrash the GC' sort of stuff so I'm not >>>> taking the results too seriously yet, but I'm still a bit confused - how >>>> can this be? What sort of voodoo is asm.js up to here? >>>> >>>> Also: does asm.js produce garbage if it's just running 'raw' code, ie: >>>> no API calls? >>>> >>>> Bye! >>>> Mark >>>> >>>> >>>> >>>> >>>> >>>> -- >>> You received this message because you are subscribed to the Google >>> Groups "emscripten-discuss" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> For more options, visit https://groups.google.com/d/optout. >>> >> >> -- >> You received this message because you are subscribed to a topic in the >> Google Groups "emscripten-discuss" group. >> To unsubscribe from this topic, visit >> https://groups.google.com/d/topic/emscripten-discuss/yXE_YRD1Zus/unsubscribe >> . >> To unsubscribe from this group and all its topics, send an email to >> [email protected]. >> For more options, visit https://groups.google.com/d/optout. >> > > -- > You received this message because you are subscribed to the Google Groups > "emscripten-discuss" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "emscripten-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
